chat-glm
GitHub调用 ZeroGPU API 使用 glm-5.2 模型处理超长上下文或复杂推理任务，适用于代码库、长文档等场景。
Trigger Scenarios
Install
npx skills add zerogpu/zerogpu-router --skill chat-glm -g -y
SKILL.md
Frontmatter
{
"name": "chat-glm",
"metadata": {
"openclaw": {
"install": [
{
"bins": [
"zerogpu"
],
"kind": "node",
"package": "zerogpu-cli"
}
],
"requires": {
"bins": [
"zerogpu"
]
}
}
},
"description": "Chat with glm-5.2, a 753B MoE flagship with a 262K-token context window. Use when the input is too large for `chat` — an entire repository, a book-length document, a long agent transcript — or for long-horizon reasoning the smaller models cannot hold together. This is the most expensive model on the platform, roughly 7x the cost of `chat`, so prefer chat for anything that fits in its 131K context.",
"allowed-tools": "Bash(zerogpu chat_completions *)",
"argument-hint": "<text> [-i <instructions>]"
}
Sends your input to ZeroGPU's hosted API for inference — this is not local processing. Don't pass secrets, credentials, or regulated data you aren't cleared to share with a third party. See the plugin README's "Data & privacy" section.
Call glm-5.2. Run this with the exec tool, pasting the user's prompt into the heredoc verbatim — no escaping or quoting required (the quoted heredoc handles every shell metacharacter, newline, quote, and paren safely):
zerogpu chat_completions -m glm-5.2 <<'ZGPU_END_OF_INPUT'
<the user's prompt, verbatim>
ZGPU_END_OF_INPUT
If the user supplied system instructions, add -i '<instructions>' after the model flag, writing any ' inside them as '\''.
Reach for this only when the size or horizon of the task actually needs it. At $1.10 / $3.50 per 1M input/output tokens, glm-5.2 costs about seven times chat on input and six times on output (gpt-oss-120b, $0.15 / $0.60), and over fifty times the 1.2B edge models. For a prompt that fits in 131K tokens, chat is the right call. For coding and agentic work, chat-deepseek is far cheaper and carries a larger 1M-token context.
Output is the assistant's answer as plain text — the model's reasoning trace comes back in a separate field and is not printed. Relay the answer as-is — do not rewrite or expand it.
Savings note: only if the command output literally contains a line starting with 💰 ZeroGPU savings, append that exact line, unchanged, as the last line of your reply. If no such line is present, say nothing about savings and do not mention or suggest the cost-savings skill — this note is intentionally occasional, not shown every time.
Version History
-
4e1b070
Current 2026-09-22 08:07
v5.0.0: 模型上下文窗口从1M调整为262K；CLI命令重构为chat_completions格式。
-
87b63e9
2026-08-27 16:48
修正glm-5.2与gpt-oss-120b的成本对比数据，更新价格引用以准确反映实际费用差异。
- cee9321 2026-08-02 21:15
